Chapter Five: Digital Audio

5. Sampling Rates

In reproducing high-quality audio, there are three possible solutions to the problem of aliasing caused by trying to represent frequencies above than the Nyquist frequency. The first is to band-limit the frequencies allowed to enter the system. The second is to have a higher sampling rate, thereby increasing the frequency range before aliasing occurs. In reality, both of these have been addressed by current digital audio systems. A third solution is to use an over-sampling sigma-delta converter which compensates for lower sample rates by using higher sample sizes, 1-bit ADC's and oversampling to maintain frequency response.

An anti-aliasing filter is placed in front of an analog-to-digital converter to prevent frequencies above the Nyquist frequency from ever being sampled. So far, existing technology has not afforded us a true "brick wall" filter that completely eliminates unwanted frequencies without having any effect on those in the "legal" range below the Nyquist frequency. These is still a rolloff curve that attenuates frequencies closest to the cutoff at a commonly used sampling rate.

cutoff filter rolloff

The diagram above indicates that even at a sampling rate of 44.1 kHz (the CD rate), some audible frequencies, pictured in red, are attenuated by the filter. How much is attenuated is determined somewhat by the quality of the filter and its steepness (or 'Q'). The higher the sampling rate used, the less noticeable is the impact of the filter rolloff on audible frequencies, because more and more of the rolloff is above audio rate. Some believe that even ultrasonic frequencies generated above human hearing, but still in the rolloff area may intermodulate with audible frequencies and cause modest distortion. This is part of the justification for digital audio workstations, such as Pro Tools and Digital Performer, giving users the option of sampling rates higher than 44.1 kHz (e.g., 96 kHz, 192 kHz). Current computer speeds, audio interface capabilities, and storage speed and size, have made this possible.

Current industry standards for full-quality (i.e., uncompressed) digital audio sample rates are:

rate Use(s)
32 kHz voice quality
44.1 kHz CD, DAT, digital recording software/hardware, some video
48 kHz video, digital recording software/hardware
96 kHz digital recording software/hardware while recording/editing
192 kHz digital recording software/hardware while recording/editing
282.24 kHz Used for audio stream of Sony/Phillips SACD (Super-audio CD), and could actually be up to 4x's as fast
Factoid: Why the 44.1K sampling rate for CD's and not 48K?

Why did the CD standard settle on 44.1K rather than say 48K? Most likely it was that video equipment already had clocks that ran at 44.1K for PAL and NTSC format 3-channel digital audio (which was actually encoded into the video signal outside the picture area) that could be easily be integrated into the first CD players. I have also heard that Herbert von Karajan complained to Sony that Beethoven's 9th Symphony would not fit on the early proposed CD specifications. By lowering the rate to 44.1K, 74 minutes could be recorded onto a CD using 16-bit samples, enough to do the trick.